EnergyAware Routing

1
Energy-Aware Routing
Robert Murawski February 5, 2008

• Paper 1 Wireless sensor networks a survey
• Paper 2 Online Power-aware Routing in
  Wireless Ad-hoc Networks

2
Energy-Aware Routing

• Paper 1
• Wireless sensor networks a survey
• Focus on Network Layer
• Energy Aware Sections
• Paper 2
• Development of Specific Energy-Aware Routing
  Algorithm
• Online Algorithm, and a Practical Implementation
  of the Algorithm

3
Paper 1 Sensor Network Survey

• Network Layer Considerations for Sensor Networks
• Power Efficiency
• Data Centric Information
• Data Aggregation
• Attribute Based Addressing / Location Awareness
• Focus of this Presentation
• 1 Power Efficiency in Sensor Network Routing

4
Energy Efficient Routing

• Approaches for Selecting an Energy-Efficient
  Route
• Maximum Available Power (PA) Route
• Select paths containing nodes with the most
  residual power
• Minimum Energy (ME) Route
• Select paths that consume the least amount of
  energy
• Minimum Hop (MH) Route
• Select paths that utilize the least amount of
  network hops
• Maximum Minimum PA Node Route
• Select the route that maximizes the minimum
  residual energy

5
Route Selection Example
PA Residual Power of Nodeai Energy Required
to Transmit a Message through Link
Source Wireless sensor networks a survey,
I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E.
Cayirci
6
Available Routes

• Four Possible Routes ( T ? Sink)
• T ? B ? A ? Sink
• PA 4, Total a 3
• T ? C ? B ? A ? Sink
• PA 6, Total a 6
• T ? D ? Sink
• PA 3, Total a 4
• T ? F ? E ? Sink
• PA 5, Total a 6

7
Route Selection

• Maximum Available Power (PA)
• Route 2 (T?C?B?A?Sink) has the largest PA Value
• Route 2 is an extension of Route 1 (T?B?A?Sink)
• Must not consider routes extended from other
  routes by adding additional nodes
• Route 4 (T?F?E?Sink) would be selected

8
Route Selection

• Minimum Energy (ME)
• Route 1 (T?B?A?Sink)
• Consumes the Least amount of energy
• Minimum Hop (MH)
• Route 3 (T?D?Sink)
• Lowest Hop Count
• Maximum Minimum PA
• Route 3 (T?D?Sink)
• Minimum PA Value 3
• Minimum PA for Other Routes 2

9
Routing Schemes for Sensor Networks

• Small Minimum Energy Communication Network
  (SMECN)
• Given a Network Graph G
• Weighted Graph G(V,E)
• Compute an Energy Efficient Subgraph G
• All nodes in G are present also in subgraph G
• Number of edges in subgraph G are less than in G
• All end-to-end node connections in G are also in
  subgraph G
• Energy required to transmit a message from node u
  to node v in subgraph G is less than the power
  required in graph G
• For every route u ? v in graph G, there is a
  Minimum Energy (ME) route u ? v in subgraph G
• Details on generating subgraph A are not Given,
  only a description of the subgraph and its use in
  optimizing energy

10
SMECN Cont.

• Transmission Power
• t is a constant
• n pathloss exponent
• d distance between u and v
• Network Path
• Power Required to route a message
• c receive power
• Path r in subgraph G is a minimum energy path if
  for all routes r in graph G
• Overview of SMECN
• Once subgraph G is computed, nodes can easily
  select the path that requires the least energy
  consumption from all available routes

11
More Energy-Efficient Routing Schemes for Sensor
Networks

• Sensor Protocols for information via negotiation
  (SPIN)
• Limit the amount of information transmitted via
  the network
• Control Message Sequence
• ADV Message
• Contains a description of the data to be sent.
• REQ Message
• If the neighbor node is interested in the data,
  send a REQ message back to the source node.
• DATA
• The source node transmits the data to nodes that
  request it.

Source Wireless sensor networks a survey,
I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E.
Cayirci
12
More Energy-Efficient Routing Schemes for Sensor
Networks

• Sensor Protocols for information via negotiation
  (SPIN)
• For sensor nodes, the ADV message contains a
  descriptor of the DATA message
• i.e. image, sensor reading descriptions
• Nodes only transmit the large DATA packets when
  necessary
• Reducing the amount of large DATA transmissions
  increases the residual energy of nodes within the
  network.

13
More Energy-Efficient Routing Schemes for Sensor
Networks

• Low-energy adaptive clustering hierarchy (LEACH)
• Reduce the energy dissipation in the network
• Backhaul of data to a base station can be costly
• Designate cluster-head nodes that backhaul
  aggregate data from all nodes within the cluster
  to the base station
• Phase 1) Setup Phase
• Sensor nodes are randomly chosen to be
  Cluster-heads
• All Cluster heads advertise to all nodes within
  the network.
• Non-cluster head nodes choose their cluster based
  on signal strength of the advertisement.
• Phase 2) Steady State Phase
• Non-cluster head nodes send data to their
  designated cluster-head
• Cluster heads backhaul information to the base
  station.

14
Paper 2

• Online Power-aware Routing in Wireless Ad-hoc
  Networks
• Papers focus
• Development of an on-line power-aware routing
  protocol
• On-line Protocol does not know the sequence of
  messages to be routed ahead of time
• max-min zPmin Algorithm
• Requires knowledge of power availability of all
  nodes within the network, impractical for large
  networks
• A second algorithm zone-based routing
• A more practical version of the max-min zPmin
  theorem

15
Introduction

• Power Consumption in Ad-hoc Networks
• Message Transmission
• Message Reception
• Node Idle Time
• Focus of this paper
• Minimizing power consumption during communication
  (transmission and reception)

16
Introduction

• Standard metrics for optimizing power-routing
• Minimize energy consumed for each message
• Minimize variance in each computer power level
• Minimize radio of cost/packet
• Minimize the maximum node cost
• Drawback of these metrics
• Focus on individual nodes, not the system as a
  whole
• Could lead to a system of nodes with high
  residual power, but with several key nodes
  depleted of power
• This paper
• Focus on maximizing the lifetime of the network
• Lifetime time to the earliest time a message
  cannot be sent

17
System Model

• Network View a weighted graph G(V,E)
• Vertices are computers within the network
• Weight of a vertex corresponds to the residual
  power of the node.
• Edges are pairs of computers within communication
  range
• Weight of an edge is the cost in power of sending
  a unit message
• Large messages are simply multiples of the unit
  message
• Power to transmit a message
• k and c are constants defined by the wireless
  technology used.

18
Max-min Path

• Intuition
• Route message over paths with the maximum minimum
  residual energy
• Find all possible paths from source to
  destination
• Determine the minimum residual energy node for
  each path
• Choose the path with the maximum residual power
  for each node
• Using the max-min path can have poor performance
  as seen in the following hypothetical network.

19
Max-min Path

• Assumptions
• Initial power for intermediate nodes 20
• Initial power for source node 8
• Weight of edge on the arc 1
• Weight of straight edge 2
• Max-min Path
• Route through the arc
• Residual Power of all nodes after one message
• (20 1) / 20 95
• 20 messages can be sent total.
• Optimal Path
• Route messages through the straight paths
• Residual Power of intermediate node after
  transmission through a straight path
• (20 2) / 20 90
• 10 (n 4) message can be sent total.
• See Right for a network of 8 nodes, 40 messages
  can be sent.
• Increases with network size

20
The z Parameter

• Two extreme solutions to power-aware routing
• Compute the path with minimal power consumption
  Pmin
• Compute the path that maximizes the minimal
  residual power
• Authors Goal Optimize for both 1 and 2
• Methodology
• Relax the Pmin requirement by a factor of z. (z
  1)
• For z 2, select a path that consumes no more
  than twice the minimum possible energy
  consumption of all possible routes
• Max-min zPmin Algorithm
• Select path that consumes at most zPmin while
  maximizing the minimal residual power fraction

21
Max-min zPmin Algorithm

• Definitions
• P(vi) Initial power level of node vi (at time
  t0)
• eij weight of the edge between vi and vj (cost
  of transmission)
• Pt(vi) Power of node vi at time t
• utij Residual power of node vi after sending
  message to node j

22
Max-min zPmin Algorithm

• 0) Find the path with the least power
  consumption, Pmin
• 1) Find the path with the least power
  consumption in the graph
• If the power consumption zPmin or no path is
  found, use the previously computed path and stop.
• 2) Find the minimal utij on the path from step 1
  (umin)
• 3) Find all edges whose residual power fraction
  utij umin and remove them from the graph.
• 4) Goto Step 1

23
Max-min zPmin

• What this algorithm accomplishes
• Step 0 First computes the Pmin
• As was shown, the Pmin can perform poorly
• Steps 1-4
• Compute the Pmin for the current state of the
  graph
• Determine if this value of Pmin is above the
  relaxed requirement of zPmin
• If the remaining graph is within bounds (zPmin),
  determine the minimum residual power for all
  nodes within the current Pmin route (umin)
• Eliminate edges within the graph that do meet or
  exceed the umin
• Note Eliminates the current Pmin path found in
  step 1
• Repeat steps 1 thru 4 for the newly updated
  version of the graph

24
Max-min zPmin

• What this algorithm accomplishes
• Iteratively finds paths with higher cost, while
  remaining below the threshold zPmin
• Eliminates edges that result in depleted residual
  power in intermediate nodes.
• Resulting path is a trade-off between the pure
  max-min path and the pure Pmin path

25
Choosing the Z Parameter

• Extremes values of Z
• Set Z to 1
• Reduces xPmin algorithm to the purely Pmin path
• Set Z to 8
• Reduces xPmin algorithm to the purely min-max
  path
• Authors Focus
• Adaptive method for computing the Z parameter
  that maximizes the network lifetime

26
Choosing the Z Parameter

• 0) Choose initial value for z, and step size d.
• 1) Run max-min zPmin algorithm for some interval
  T
• 2) Compute for all hosts, let the minimal
  be t1
• 3) Increase z by d, run max-min zPmin for
  interval T
• 4) Compute for all hosts, let the minimal be
  t2
• 5) If any host is saturated, exit (use this
  value of z)
• 6) If t1
• 7) If t1 t2, set d - d /2, t1 t2, and goto
  step 3

27
Results for Max-min zPmin

• Authors Claim
• Results show adaptively selecting z leads to
  superior performance over the minimal power
  algorithm (z1) and the max-min algorithm (z8)
• Question
• The results do show better results than the
  max-min algorithm.
• The results show little or no improvement over
  the Pmin algorithm
• When z1, the maximum (or near maximum) value
  is achieved.
• Better resolution graph may be necessary.

28
Zone-Based Routing

• The max-min zPmin algorithm is hard to implement
  on large scale networks
• Accurate knowledge for all node available power
  is required.
• In large networks would result in large control
  overhead, defeating the purpose of
  energy-efficient routing
• A hierarchical approach to the max-min zPmin
  algorithm
• Group nodes into zones (based on geographic
  positioning)
• Zone hosts direct local routing
• Messages are routed through zones based on the
  power of the zone.
• Issues to consider
• How zone hosts estimate the power of the zone
• How to route messages within a zone
• How to route messages between zones

29
Zone-Based Routing

• Zone Power Estimation
• Zone Power Estimate of of messages that can
  flow through the zone
• Estimation is relative to the direction of the
  message transition
• Zone assumed to be square
• neighbors north, south, east, west
• Neighbor zones overlap
• Estimation Process
• Choose ?, and P 0
• Repeat
• Find max-min zPmin for ? Messages
• Send ? Message through the zone
• P P ?
• (Until some nodes are saturated)
• Return P

30
Zone-Based Routing

• Global Path Selection
• View the Global network as a weighted graph
• Each zone has 5 vertices (for square zones)
• 1. Zone, and 4. Directions
• Zone Weight Infinite
• Direction Weight Power level for sending
  message through in this direction
• From Previous Slide
• There are no edge weights
• To Route between zones, us the max-min algorithm
  on the zone graph
• Bias routing for zones with higher power levels
  (modified Bellman-Ford)

31
Zone-Based Routing

• Local paths are determined using the max-min Pmin
  Algorithm
• To select Zone-edge messages
• There can be multiple nodes within a zone edge
• Zone Edge Overlap between two zones
• Example Route messages between A and C (B in
  the middle)
• Select the highest weight node in the section AB
• Select the highest weight node in the section BC
• Use min-max zPmin algorithm to compute the best
  path between the edge nodes.

32
Zone-Based Routing

• Results
• Zone-Base Routing vs. max-min zPmin
• Requires Less Control Overhead
• Sacrifices Overall Performance
• Two Scenarios
• 94.5 and 96 lifetime of the max-min zPmin is
  achieved
• Max-min zPmin 1000 control messages flooded
  (1000 nodes)
• Zone-Based 24 control messages flooded (24
  zones)
• After Zone Power Estimation
• 42 Nodes per Zone (assuming even distribution)
• Zone-based routing dramatically reduces the
  simulation running time

33
Conclusion

• First Paper
• Overview of power-aware routing
• Second Paper
• Two algorithms developed for power-aware routing
• Max-min zPmin More effective at the cost of
  large overhead
• Zone-based 95 of max-min zPmin is achieved
  with significantly less overhead

34
Thank you!

• Question/Comments?